ON TARGET

SCROLL

The new site selection criteria

Data center construction has changed

While I have written about site selection in past columns, I find it is time to revisit the topic because of the vast changes in criteria. What was happening prior no longer applies.

Site selection criteria used to require around 15 acres (or so) with about 40 MW per site. The new criteria call for 200-plus acres and 150 MW per site. This change is obviously due to the needs of the hyperscale community. Across the entire enterprise, processing is absorbed by colo and web services, with wholesale providers turning to leasing to meet hyperscale demand. In the past, I wrote about the “Industrial Age” of building hyperscale buildings, yet wholesale providers have not adjusted to designing metal buildings or other criteria.

Land criteria — While land sizes have become much bigger, there are other factors to consider than just the amount of land. Much of the new land at these acreages is farmland, which has been handed down from generation to generation. Whereas previously, sites at 15 acres were preengineered for data center use, the newer farmland is not, and environmental and civil engineering is needed. Typically, we have seen soil borings that hit bedrock between 6 inches and 20 feet, leaving a wide variety of options during borings.

Price  In the past, when we provided data center development services (master planning), the typical market fee was around $30,000. Now, with preengineering costs, environmental fees, and utility coordination expenses, the total climbs to over $400,000 per site — a significant difference from before. The project management process is still around $30,000, while the other consultants balance out the remainder. The other factor that goes into pricing is the coordination with local city authorities. Site plans are developed and presented to the city, which may or may not be supported by local government representatives. Remember that many of the new hyperscale projects are not “pretty” and may very much resemble a manufacturing facility.

“In the past, when we provided data center development services (master planning), the typical market fee was around $30,000. Now, with preengineering costs, environmental fees, and utility coordination expenses, the total climbs to over $400,000 per site. ”

Master Planning — There are multiple factors that go into master planning. First, are the renderings submitted to the local authorities. One of the challenges wholesale developers have is that each hyperscale client has its own unique prototype, and the renderings are not pretty. Instead of a true master plan, they prefer more of a “masking plan” identifying the total capability of the land buildout. The initial submittal becomes difficult on a city-by-city basis. This, combined with environmental/geotechnical and civil engineering items as well as utilities, makes the initial masking plan more difficult and expensive.

Utility planning — Finding a 40- to 45-MW site has proven challenging, and, now, most hyperscalers are looking for sites that are 150-plus MW per site —all at a reasonable utility rate with a promising mechanical environment. Utility companies typically have some capacity. However, in many cases, to get to a hyperscale installation, they will have to build additional capacity. Transformers (generally from China) take 12 to 18 months to procure and install. In a phased masking plan, we can work through this issue; however, if it's not phased, problems can occur. Additionally, downstream electrical configuration matters. Many of the hyperscalers require five-nines availability (99.999 %). To achieve this in a catcher-block system, you will need to acquire two redundant feeds. Within a distributed, redundant scheme, only one feed to achieve five-nines availability is required. Also, since a distributed architecture requires “chuncks” of power versus the more scalable catcher-block system, its overall total cost of ownership when completely installed will be the lowest at the end of the day.

Conclusion

We work in a very dynamic industry. While some things do not change much, such as UPSs and generators, things like locations and processing do, and it's very important for us as engineers to keep on top of these items. We are also in the era of retrofits. The average life span of a UPS is 15 years. That means upgrades will be needed soon, so there's plenty of work for all of us.


Dress shirt, Forehead, Smile, Chin, Jaw, Sleeve, Collar

Paul Schlattman

Paul Schlattman is practice leader of mission critical facilities for Environmental Systems Design Inc. Schlattman is also a member of Mission Critical’s technical advisory board.